As a practicing psychiatrist, I am faced with making decisions for my patients to increase the likelihood of good outcomes and decrease the probabilities of harmful adverse effects. When I make these decisions, I have found all too often that data comparing treatment options are lacking. Few studies have been done that directly compare treatments. As a result, I have to “guestimate” which treatment might be best—but I am still left with too much uncertainty. I cannot use results of clinical trials from different studies because each study has a different population and different placebo response rates. For example, if one study showed that a medication resulted in 70% response and another study showed 50% response, I can't assume that the first medication is better if the response rates for placebo for the first study was 50% and the second study 30%. In the absence of direct comparisons, are there any alternatives that might inform our decisions?
One attractive alternative to consider are the statistics number needed to treat (NNT), number needed to harm (NNH), and the likelihood to be helped or harmed (LHH).1–3 On the face of it, the NNT is not intuitive—it is “the number of patients you would need to treat with one medication instead of another intervention (such as placebo) before you would expect to encounter an additional positive outcome of interest (such as response or remission).”4 The lower the NNT, the fewer the number of patients needed to encounter an additional positive outcome, the more effective the intervention. Conversely, the higher the NNH, the less likely the adverse event. The LHH ratio of the NNT/NNH can be interpreted as the higher the number, the more likely patients are to be helped rather than harmed, and this ratio can be calculated for the harm of individual side effects.
Although inexact and imperfect, NNT, NNH, and LHH can at least provide an estimate of the relative strengths and weaknesses of competing treatments. Clinicians can use these tools to inform their clinical decision-making in the absence of real-world comparative effectiveness studies.
- Citrome L. Number needed to treat: what it is and what it isn't, and why every clinician should know how to calculate it. J Clin Psychiatry. 2011;72(3):412–413. doi:10.4088/JCP.11ac06874 [CrossRef] PMID:21450157
- Citrome L, Ketter TA. Number needed to treat can be helpful: a response to Alphs et al. Eur Neuropsychopharmacol. 2013;23(11):1656–1657. doi:10.1016/j.euroneuro.2013.02.008 [CrossRef] PMID:23712091
- Citrome L, Ketter TA. When does a difference make a difference? Interpretation of number needed to treat, number needed to harm, and likelihood to be helped or harmed. Int J Clin Pract. 2013;67(5):407–411. doi:10.1111/ijcp.12142 [CrossRef] PMID:23574101
- Citrome L. Food and Drug Administration-approved treatments for acute bipolar depression: what we have and what we need. J Clin Psychopharmacol. 2020;40(4):334–338. doi:10.1097/JCP.0000000000001227 [CrossRef] PMID:32639285